Results for “petascale computing” — Page 9

DOE OSTI2014

A Pervasive Parallel Processing Framework for Data Visualization and Analysis at Extreme Scale

The evolution of the computing world from teraflop to petaflop has been relatively effortless, with several of the existing programming models scaling effectively to the petascale. The migration to exascale, however, poses considerable challenges. All industry trends infer that the exascale machine will be built using processors containing hundreds to thousands of cores per chip. It can be inferred that efficient concurrency on exascale machines requires a massive amount of concurrent threads, each performing many operations on a localized piece of data. Currently, visualization libraries and applications are based off what is known as the visualization pipeline. In the pipeline model, algorithms are encapsulated as filters with inputs and outputs. These filters are connected by setting the output of one component to the input of another. Parallelism in the visualization pipeline is achieved by replicating the pipeline for each processing thread. This works well for today’s distributed memory parallel computers but cannot be sustained when operating on processors with thousands of cores. Our project investigates a new visualization framework designed to exhibit the pervasive parallelism necessary for extreme scale machines. Our framework achieves this by defining algorithms in terms of worklets, which are localized stateless operations. Worklets are atomic operations that execute when invoked unlike filters, which execute when a pipeline request occurs. The worklet design allows execution on a massive amount of lightweight threads with minimal overhead. Only with such fine-grained parallelism can we hope to fill the billions of threads we expect will be necessary for efficient computation on an exascale machine.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2020

Characterization and identification of HPC applications at leadership computing facility

High Performance Computing (HPC) is an important method for scientific discovery via large-scale simulation, data analysis, or artificial intelligence. Leadership-class supercomputers are expensive, but essential to run large HPC applications. The Petascale era of supercomputers began in 2008, with the first machines achieving performance in excess of one petaflops, and with the advent of new supercomputers in 2021 (e.g., Aurora, Frontier), the Exascale era will soon begin. However, the high theoretical computing capability (i.e., peak FLOPS) of a machine is not the only meaningful target when designing a supercomputer, as the resources demand of applications varies. A deep understanding of the characterization of applications that run on a leadership supercomputer is one of the most important ways for planning its design, development and operation. In order to improve our understanding of HPC applications, user demands and resource usage characteristics, we perform correlative analysis of various logs for different subsystems of a leadership supercomputer. This analysis reveals surprising, sometimes counter-intuitive patterns, which, in some cases, conflicts with existing assumptions, and have important implications for future system designs as well as supercomputer operations. For example, our analysis shows that while the applications spend significant time on MPI, most applications spend very little time on file I/O. Combined analysis of hardware event logs and task failure logs show that the probability of a hardware FATAL event causing task failure is low. Combined analysis of control system logs and file I/O logs reveals that pure POSIX I/O is used more widely than higher level parallel I/O. Based on holistic insights of the application gained through combined and co-analysis of multiple logs from different perspectives and general intuition, we engineer features to "fingerprint" HPC applications. We use t-SNE (a machine learning technique for dimensionality reduction) to validate the explainability of our features and finally train machine learning models to identify HPC applications or group those with similar characteristic. To the best of our knowledge, this is the first work that combines logs on file I/O, computing, and inter-node communication for insightful analysis of HPC applications in production.

Liu, Zhengchun↗

DOE OSTI2013

PoPLAR: Portal for Petascale Lifescience Applications and Research

Background: We are focusing specifically on fast data analysis and retrieval in bioinformatics that will have a direct impact on the quality of human health and the environment. The exponential growth of data generated in biology research, from small atoms to big ecosystems, necessitates an increasingly large computational component to perform analyses. Novel DNA sequencing technologies and complementary high-throughput approaches–such as proteomics, genomics, metabolomics, and meta-genomics–drive data-intensive bioinformatics. While individual research centers or universities could once provide for these applications, this is no longer the case. Today, only specialized national centers can deliver the level of computing resources required to meet the challenges posed by rapid data growth and the resulting computational demand. Consequently, we are developing massively parallel applications to analyze the growing flood of biological data and contribute to the rapid discovery of novel knowledge. Methods: The efforts of previous National Science Foundation (NSF) projects provided for the generation of parallel modules for widely used bioinformatics applications on the Kraken supercomputer. We have profiled and optimized the code of some of the scientific community’s most widely used desktop and small-cluster-based applications, including BLAST from the National Center for Biotechnology Information (NCBI), HMMER, and MUSCLE; scaled them to tens of thousands of cores on high-performance computing (HPC) architectures; made them robust and portable to next-generation architectures; and incorporated these parallel applications in science gateways with a web-based portal. Results: This paper will discuss the various developmental stages, challenges, and solutions involved in taking bioinformatics applications from the desktop to petascale with a front-end portal for very-large-scale data analysis in the life sciences. Conclusions: This research will help to bridge the gap between the rate of data generation and the speed at which scientists can study this data. The ability to rapidly analyze data at such a large scale is having a significant, direct impact on science achieved by collaborators who are currently using these tools on supercomputers.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2016

A low-communication, parallel algorithm for solving PDEs based on range decomposition

This work proposes a new, low-communication algorithm for solving PDEs on massively parallel computers. The range decomposition (RD) algorithm exposes coarse-grain parallelism by applying nested iteration and adaptive mesh refinement locally before performing a global communication step. Just a few such steps are observed to be sufficient to obtain accuracy within a small multiple of discretization error. The target applications are petascale and exascale machines, where hierarchical parallelism is required and traditional parallel numerical PDE communication patterns are costly because of message latency. The RD algorithm uses a partition of unity to equally distribute the error, and thus, the work. The computational advantages of this approach are that the decomposed problems can be solved in parallel without any communication until the partitioned solutions are summed. This offers potential advantages in the paradigm of expensive communication but very cheap computation. This paper introduces the method and explains the details of the communication step. Two performance models are developed, showing that the latency cost associated with a traditional parallel implementation of nested iteration is proportional to log(P) 2 , whereas the RD method reduces the communication latency to log(P), while maintaining similar bandwidth costs. Numerical results for two problems, Laplace and advection diffusion, demonstrate the enhanced performance, and a heuristic argument explains why the method converges quickly.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2016

Analyzing How We Do Analysis and Consume Data, Results from the SciDAC-Data Project

One of the main goals of the Dept. of Energy funded SciDAC-Data project is to analyze the more than 410,000 high energy physics datasets that have been collected, generated and defined over the past two decades by experiments using the Fermilab storage facilities. These datasets have been used as the input to over 5.6 million recorded analysis projects, for which detailed analytics have been gathered. The analytics and meta information for these datasets and analysis projects are being combined with knowledge of their part of the HEP analysis chains for major experiments to understand how modern computing and data delivery is being used. We present the first results of this project, which examine in detail how the CDF, D0, NOvA, MINERvA and MicroBooNE experiments have organized, classified and consumed petascale datasets to produce their physics results. The results include analysis of the correlations in dataset/file overlap, data usage patterns, data popularity, dataset dependency and temporary dataset consumption. The results provide critical insight into how workflows and data delivery schemes can be combined with different caching strategies to more efficiently perform the work required to mine these large HEP data volumes and to understand the physics analysis requirements for the next generation of HEP computing facilities. In particular we present a detailed analysis of the NOvA data organization and consumption model corresponding to their first and second oscillation results (2014-2016) and the first look at the analysis of the Tevatron Run II experiments. We present statistical distributions for the characterization of these data and data driven models describing their consumption

Ding, P.↗

DOE OSTI2018

Autotuning in High-Performance Computing Applications

Autotuning refers to the automatic generation of a search space of possible implementations of a computation that are evaluated through models and/or empirical measurement to identify the most desirable implementation. Autotuning has the potential to dramatically improve the performance portability of petascale and exascale applications. To date, autotuning has been used primarily in high-performance applications through tunable libraries or previously tuned application code that is integrated directly into the application. This paper draws on the authors' extensive experience applying autotuning to high-performance applications, describing both successes and future challenges. If autotuning is to be widely used in the HPC community, researchers must address the software engineering challenges, manage configuration overheads, and continue to demonstrate significant performance gains and portability across architectures. In particular, tools that configure the application must be integrated into the application build process so that tuning can be reapplied as the application and target architectures evolve.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2012

Argonne Leadership Computing Facility 2011 annual report: Shaping future supercomputing

The ALCF's Early Science Program aims to prepare key applications for the architecture and scale of Mira and to solidify libraries and infrastructure that will pave the way for other future production applications. Two billion core-hours have been allocated to 16 Early Science projects on Mira. The projects, in addition to promising delivery of exciting new science, are all based on state-of-the-art, petascale, parallel applications. The project teams, in collaboration with ALCF staff and IBM, have undertaken intensive efforts to adapt their software to take advantage of Mira's Blue Gene/Q architecture, which, in a number of ways, is a precursor to future high-performance-computing architecture. The Argonne Leadership Computing Facility (ALCF) enables transformative science that solves some of the most difficult challenges in biology, chemistry, energy, climate, materials, physics, and other scientific realms. Users partnering with ALCF staff have reached research milestones previously unattainable, due to the ALCF's world-class supercomputing resources and expertise in computation science. In 2011, the ALCF's commitment to providing outstanding science and leadership-class resources was honored with several prestigious awards. Research on multiscale brain blood flow simulations was named a Gordon Bell Prize finalist. Intrepid, the ALCF's BG/P system, ranked No. 1 on the Graph 500 list for the second consecutive year. The next-generation BG/Q prototype again topped the Green500 list. Skilled experts at the ALCF enable researchers to conduct breakthrough science on the Blue Gene system in key ways. The Catalyst Team matches project PIs with experienced computational scientists to maximize and accelerate research in their specific scientific domains. The Performance Engineering Team facilitates the effective use of applications on the Blue Gene system by assessing and improving the algorithms used by applications and the techniques used to implement those algorithms. The Data Analytics and Visualization Team lends expertise in tools and methods for high-performance, post-processing of large datasets, interactive data exploration, batch visualization, and production visualization. The Operations Team ensures that system hardware and software work reliably and optimally; system tools are matched to the unique system architectures and scale of ALCF resources; the entire system software stack works smoothly together; and I/O performance issues, bug fixes, and requests for system software are addressed. The User Services and Outreach Team offers frontline services and support to existing and potential ALCF users. The team also provides marketing and outreach to users, DOE, and the broader community.

99 GENERAL AND MISCELLANEOUS↗

DOE OSTI2018

Deploy threading in Nalu solver stack

The goal of the ExaWind project is to enable predictive simulations of wind farms composed of many MW-scale turbines situated in complex terrain. Predictive simulations will require computational fluid dynamics (CFD) simulations for which the mesh resolves the geometry of the turbines, and captures the rotation and large deflections of blades. Whereas such simulations for a single turbine are arguably petascale class, multi-turbine wind farm simulations will require exascale-class resources. The primary code in the ExaWind project is Nalu, which is an unstructured-grid solver for the acousticallyincompressible Navier-Stokes equations, and mass continuity is maintained through pressure projection. The model consists of the mass-continuity Poisson-type equation for pressure and a momentum equation for the velocity. For such modeling approaches, simulation times are dominated by linear-system setup and solution for the continuity and momentum systems. For the ExaWind challenge problem, the moving meshes greatly affect overall solver costs as re-initialization of matrices and re-computation of preconditioners is required at every time step In this Milestone, we examine the effect of threading on the solver stack performance against flat-MPI results obtained from previous milestones using Haswell performance data full-turbine simulations. Whereas the momentum equations are solved only with the Trilinos solvers, we investigate two algebraic-multigrid preconditioners for the continuity equations: Trilinos/Muelu and HYPRE/BoomerAMG. These two packages embody smoothed-aggregation and classical Ruge-Stiiben AMG methods, respectively. In our FY18 Q2 report, we described our efforts to improve setup and solve of the continuity equations under flat-MPI parallelism. While significant improvement was demonstrated in the solve phase, setup times remained larger than expected. Starting with the optimized settings described in the Q2 report, we explore here simulation performance where OpenMP threading is employed in the solver stack. For Trilinos, threading is acheived through the Kokkos abstraction where, whereas HYPRE/BoomerAMG employs straight OpenMP. We examined results for our mid-resolution baseline turbine simulation configuration (229M DOF). Simulations on 2048 Haswell cores explored the effect of decreasing the number of MPI ranks while increasing the number of threads. Both HYPRE and Trilinos exhibited similar overal solution times, and both showed dramatic increases in simulation time in the shift from MPI ranks to OpenMP threads. This increase is attributed to the large amount of work per MPI rank starting at the single-thread configuration. Decreasing MPI ranks, while increasing threads, may be increasing simulation time due to thread synchronization and start-up overhead contributing to the latency and serial time in the model. These result showed that an MPI+OpenMP parallel decomposition will be more effective as the amount per MPI rank computation per MPI rank decreases and the communication latency increases. This idea was demonstrated in a strong scaling study of our low-resolution baseline model (29M DOF) with the Trilinos-HYPRE configuration. While MPI-only results showed scaling improvement out to about 1536 cores, engaging threading carried scaling improvements out to 4128 cores — roughly 7000 DOF per core. This is an important result as improved strong scaling is needed for simulations to be executed over sufficiently long simulated durations (i.e., for many timesteps). In addition to threading work described above, the team examined solver-performance improvements by exploring communication-overhead in the HYPRE-GMRES implementation through a communicationoptimal- GMRE algorithm (CO-GMRES), and offloading compute-intensive solver actions to GPUs. To those ends, a HYPRE mini-app was allow us to easily test different solver approaches and HYPRE parameter settings without running the entire Nalu code. With GPU acceleration on the Summitdev supercomputer, a 20x speedup was achieved for the overall preconditioner and solver execution time for the mini-app. A study on Haswell processors showed that CO-GMRES provides benefits as one increases MPI ranks.

17 WIND ENERGY↗

DOE OSTI2018

Decrease time-to-solution through improved linear-system setup and solve

The goal of the ExaWind project is to enable predictive simulations of wind farms composed of many MW-scale turbines situated in complex terrain. Predictive simulations will require computational fluid dynamics (CFD) simulations for which the mesh resolves the geometry of the turbines, and captures the rotation and large deflections of blades. Whereas such simulations for a single turbine are arguably petascale class, multi-turbine wind farm simulations will require exascale-class resources. The primary code in the ExaWind project is Nalu, which is an unstructured-grid solver for the acoustically-incompressible Navier-Stokes equations, and mass continuity is maintained through pressure projection. The model consists of the mass-continuity Poisson-type equation for pressure and a momentum equation for the velocity. For such modeling approaches, simulation times are dominated by linear-system setup and solution for the continuity and momentum systems. For the ExaWind challenge problem, the moving meshes greatly affect overall solver costs as re-initialization of matrices and re-computation of preconditioners is required at every time step We describe in this report our efforts to decrease the setup and solution time for the mass-continuity Poisson system with respect to the benchmark timing results reported in FY18 Q1. In particular, we investigate improving and evaluating two types of algebraic multigrid (AMG) preconditioners: Classical Ruge-Stfiben AMG (C-AMG) and smoothed-aggregation AMG (SA-AMG), which are implemented in the Hypre and Trilinos/MueLu software stacks, respectively. Preconditioner performance was optimized through existing capabilities and settings.

17 WIND ENERGY↗

DOE OSTI2018

Deploy threading in Nalu solver stack

The goal of the ExaWind project is to enable predictive simulations of wind farms composed of many MW-scale turbines situated in complex terrain. Predictive simulations will require computational fluid dynamics (CFD) simulations for which the mesh resolves the geometry of the turbines, and captures the rotation and large deflections of blades. Whereas such simulations for a single turbine are arguably petascale class, multi-turbine wind farm simulations will require exascale-class resources. The primary code in the ExaWind project is Nalu, which is an unstructured-grid solver for the acoustically-incompressible Navier-Stokes equations, and mass continuity is maintained through pressure projection. The model consists of the mass-continuity Poisson-type equation for pressure and a momentum equation for the velocity. For such modeling approaches, simulation times are dominated by linear-system setup and solution for the continuity and momentum systems. For the ExaWind challenge problem, the moving meshes greatly affect overall solver costs as re-initialization of matrices and re-computation of preconditioners is required at every time step.

17 WIND ENERGY↗

DOE OSTI2019

SDN-NGenIA, a Software Defined Next Generation Integrated Architecture for HEP and Data Intensive Science

This project aims to enable the LHC and other leading programs in high energy physics and other global science domains funded by the DOE to operate with a new level of efficiency and control, through the development of a Next Generation Integrated Architecture (NGenIA) based on intelligent software defined network (SDN)-driven network systems coupled to high throughput applications. While the initial focus is on the challenging LHC use case, the systems and products being developed are general, and apply to many fields of data intensive science ranging from astrophysical sky surveys to bioinformatics and earth observation, as well as other organizations facing the challenges of extracting knowledge from distributed multi-petabyte data stores. A central concept in this development program is a new paradigm ‘consistent network operations’ among widely distributed computing and storage facilities, where stable high throughput flows, at set rates, cross load-balanced network paths, up to flexible high water marks that are adjusted in real time to accommodate other network traffic. The large smooth flows are launched and managed by SDN services that act in concert with the experiments’ site-resident data distribution and management systems, to meet the expanding needs of the science programs. The technical goals include the construction of autonomous, intelligent site-resident services that interact dynamically with network-resident services, and with the science programs’ principal data distribution and management tools, to request or command network resources in support of high throughput petascale to exascale workflow Specific work items include: 1. Developing compact Data Transfer Nodes (DTNs) with auto-tuning functions that support data transfer rates in the 100 Gbps to the 1 Tbps range when used with high throughput data transfer applications 2. Deep site orchestration among virtualized clusters, storage subsystems and subnets to successfully coschedule CPU, storage and network resources 3. Science-program designed site architectures, operational modes, and policy and resource usage priorities, adjudicated across multiple network domains and virtual organizations, using the orchestration functions and methods 4. Seamlessly extending end-to-end operation across both extra-site and intra-site boundaries through the use of Open vSwitch (OVS), FireQoS + next generation Science DMZs 5. Novel methods of system integration that enable granular control of extreme scale long distance transfers through flow matching of scattered source-destination address pairs to multi-domain dynamic circuits 6. Funneling massive sets of streams to DTNs at the site edge hosting petascale buffer pools configured for flows of 100 Gbps and up, exploiting state of the art data transfers 7. Adaptive scheduling based on pervasive end-to-end monitoring, including DTN or compute-node resident agents providing comprehensive end-system profiling 8. Developing unsupervised and supervised machine learning and modeling methods to optimize the workflow involving terabyte to multi-petabyte datasets.

47 OTHER INSTRUMENTATION↗

DOE OSTI2019

Blade-Resolved, Single-Turbine Simulations Under Atmospheric Flow

The ultimate goal of the ExaWind project is to enable scientific discovery through predictive simulations of wind farms comprised of many megawatt-scale turbines situated in complex terrain. Predictive simulations will require computational fluid dynamics (CFD) simulations for which the mesh resolves the geometry of the turbines and captures the rotation and large deflections of blades. Whereas such simulations for a single turbine are arguably petascale class, multi-turbine wind farm simulations will require exascale-class resources. The objective of this milestone is to demonstrate a 'production run' of a megawatt-scale turbine to exercise and demonstrate the turbine simulation capabilities that have been developed during the previous two years of the ExaWind project. Demonstrating this simulation capability is the critical first step in enabling Nalu-Wind to be used as a predictive modeling tool that can advance the scientific understanding of wind turbine performance. In Section 2, the milestone description is provided. Next, in Section 3, we present the NREL 5-megawatt (MW) reference turbine that was used to demonstrate Nalu-Wind's capabilities and describe the four computational meshes that were generated. In Section 4, we present results from simulations of the turbine operating under uniform inflow and homogeneous isotropic turbulence (HIT) conditions. Finally, in Section 5, notable achievements, conclusions, and a path forward are described.

17 WIND ENERGY↗

DOE OSTI2016

Petascale Orbital-Free Density Functional Theory Enabled by Small-Box Algorithms

Orbital-free density functional theory (OFDFT) is a quantum-mechanics-based method that utilizes electron density as its sole variable. The main computational cost in OFDFT is the ubiquitous use of the fast Fourier transform (FFT), which is mainly adopted to evaluate the kinetic energy density functional (KEDF) and electron–electron Coulomb interaction terms. We design and implement a small-box FFT (SBFFT) algorithm to overcome the parallelization limitations of conventional FFT algorithms. We also propose real-space truncation of the nonlocal Wang–Teter KEDF kernel. The scalability of the SBFFT is demonstrated by efficiently simulating one full optimization step (electron density, energies, forces, and stresses) of 1,024,000 lithium (Li) atoms on up to 65,536 cores. We perform other tests using Li as a test material, including calculations of physical properties of different phases of bulk Li, geometry optimizations of nanocrystalline Li, and molecular dynamics simulations of liquid Li. All of the tests yield excellent agreement with the original OFDFT results, suggesting that the OFDFT-SBFFT algorithm opens the door to efficient first-principles simulations of materials containing millions of atoms.

Chemistry↗

DOE OSTI2022

Demonstrate multi-turbine simulation with hybrid-structured / unstructured-moving-grid software stack running primarily on GPUs and propose improvements for successful KPP-2

The goal of the ExaWind project is to enable predictive simulations of wind farms comprised of many megawatt-scale turbines situated in complex terrain. Predictive simulations will require computational fluid dynamics (CFD) simulations for which the mesh resolves the geometry of the turbines, capturing the thin boundary layers, and captures the rotation and large deflections of blades. Whereas such simulations for a single turbine are arguably petascale class, multi-turbine wind farm simulations will require exascale-class resources.

17 WIND ENERGY↗

DOE OSTI2021

Demonstrate moving-grid multi-turbine simulations primarily run on GPUs and propose improvements for successful KPP-2

The goal of the ExaWind project is to enable predictive simulations of wind farms comprised of many megawatt-scale turbines situated in complex terrain. Predictive simulations will require computational fluid dynamics (CFD) simulations for which the mesh resolves the geometry of the turbines, capturing the thin boundary layers, and captures the rotation and large deflections of blades. Whereas such simulations for a single turbine are arguably petascale class, multi-turbine wind farm simulations will require exascale-class resources.

17 WIND ENERGY↗

DOE OSTI2014

Lightweight and Statistical Techniques for Petascale PetaScale Debugging

This project investigated novel techniques for debugging scientific applications on petascale architectures. In particular, we developed lightweight tools that narrow the problem space when bugs are encountered. We also developed techniques that either limit the number of tasks and the code regions to which a developer must apply a traditional debugger or that apply statistical techniques to provide direct suggestions of the location and type of error. We extend previous work on the Stack Trace Analysis Tool (STAT), that has already demonstrated scalability to over one hundred thousand MPI tasks. We also extended statistical techniques developed to isolate programming errors in widely used sequential or threaded applications in the Cooperative Bug Isolation (CBI) project to large scale parallel applications. Overall, our research substantially improved productivity on petascale platforms through a tool set for debugging that complements existing commercial tools.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2020

Demonstration and performance testing of extreme-resolution simulations with static meshes on Summit (CPU & GPU) for a parked-turbine configuration and an actuator-line (mid-fidelity model) wind farm configuration (ECP-Q4 FY2020 Milestone Report)

The goal of the ExaWind project is to enable predictive simulations of wind farms comprised of many megawatt-scale turbines situated in complex terrain. Predictive simulations will require computational fluid dynamics (CFD) simulations for which the mesh resolves the geometry of the turbines and captures the rotation and large deflections of blades. Whereas such simulations for a single turbine are arguably petascale class, multi-turbine wind farm simulations will require exascale-class resources. The primary physics codes in the ExaWind simulation environment are Nalu-Wind, an unstructured-grid solver for the acoustically incompressible Navier-Stokes equations, AMR-Wind, a block-structured-grid solver with adaptive mesh refinement capabilities, and OpenFAST, a wind-turbine structural dynamics solver. The Nalu-Wind model consists of the mass-continuity Poisson-type equation for pressure and Helmholtz-type equations for transport of momentum and other scalars. For such modeling approaches, simulation times are dominated by linear-system setup and solution for the continuity and momentum systems. For the ExaWind challenge problem, the moving meshes greatly affect overall solver costs as reinitialization of matrices and recomputation of preconditioners is required at every time step. The choice of overset-mesh methodology to model the moving and non-moving parts of the computational domain introduces constraint equations in the elliptic pressure-Poisson solver. The presence of constraints greatly affects the performance of algebraic multigrid preconditioners.

17 WIND ENERGY↗

DOE OSTI2011

Adapting Wave-front Algorithms to Efficiently Utilize Systems with Deep Communication Hierarchies

Large-scale systems increasingly exhibit a differential between intra-chip and inter-chip communication performance especially in hybrid systems using accelerators. Processorcores on the same socket are able to communicate at lower latencies, and with higher bandwidths, than cores on different sockets either within the same node or between nodes. A key challenge is to efficiently use this communication hierarchy and hence optimize performance. We consider here the class of applications that contains wavefront processing. In these applications data can only be processed after their upstream neighbors have been processed. Similar dependencies result between processors in which communication is required to pass boundary data downstream and whose cost is typically impacted by the slowest communication channel in use. In this work we develop a novel hierarchical wave-front approach that reduces the use of slower communications in the hierarchy but at the cost of additional steps in the parallel computation and higher use of on-chip communications. This tradeoff is explored using a performance model. An implementation using the Reverse-acceleration programming model on the petascale Roadrunner system demonstrates a 27% performance improvement at full system-scale on a kernel application. The approach is generally applicable to large-scale multi-core and accelerated systems where a differential in system communication performance exists.

43 PARTICLE ACCELERATORS↗